for k, v in data_pca.iteritems():
data_pca[k] = (v + meanshape) - mean
for k, v in data_patches.iteritems():
data_patches[k] = (v + meanshape) - mean
# TODO: add values so that model is equally as wide/high as cropped images
# in our case: 170 x 178, i.e. add 85 to x-vals, and 89 to y-vals
meanshape = ((meanshape - mean) + [cropsize[0] / 2, cropsize[1] / 2])
if buildPatches:
# weights for eyes and mouth
from buildlib.me_weights import weights
# build patch model
patchModel = build_patches(data_patches,
gradient=True,
lbp=True,
weights=weights,
optimize_params=True)
#patchModel = build_patches(data_patches)
# store the model
model = {}
if buildPatches:
model['patchModel'] = patchModel
model['patchModel']['canvasSize'] = [cropsize[0], cropsize[1]]
if buildScoring:
model['scoring'] = scoring
model['shapeModel'] = {}
model['shapeModel']['eigenVectors'] = eigenVectors.T.tolist()
model['shapeModel']['eigenValues'] = eigenValues
model['shapeModel']['meanShape'] = meanshape.tolist()
for k,v in data_pca.iteritems():
data_pca[k] = (v+meanshape)-mean
for k,v in data_patches.iteritems():
data_patches[k] = (v+meanshape)-mean
# TODO: add values so that model is equally as wide/high as cropped images
# in our case: 170 x 178, i.e. add 85 to x-vals, and 89 to y-vals
meanshape = ((meanshape-mean)+[cropsize[0]/2,cropsize[1]/2])
if buildPatches:
# weights for eyes and mouth
from buildlib.me_weights import weights
# build patch model
patchModel = build_patches(data_patches, gradient=True, lbp=True, weights=weights, optimize_params=True)
#patchModel = build_patches(data_patches)
# store the model
model = {}
if buildPatches:
model['patchModel'] = patchModel
model['patchModel']['canvasSize'] = [cropsize[0],cropsize[1]]
if buildScoring:
model['scoring'] = scoring
model['shapeModel'] = {}
model['shapeModel']['eigenVectors'] = eigenVectors.T.tolist()
model['shapeModel']['eigenValues'] = eigenValues
model['shapeModel']['meanShape'] = meanshape.tolist()
model['shapeModel']['numEvalues'] = len(eigenValues)
model['shapeModel']['numPtsPerSample'] = meanshape.shape[0]
if buildScoring:
scoring = getScoring(data_pca, meanshape-mean)
for k,v in data_pca.iteritems():
data_pca[k] = (v+meanshape)-mean
for k,v in data_patches.iteritems():
data_patches[k] = (v+meanshape)-mean
# TODO: add values so that model is equally as wide/high as cropped images
# in our case: 170 x 178, i.e. add 85 to x-vals, and 89 to y-vals
meanshape = ((meanshape-mean)+[cropsize[0]/2,cropsize[1]/2])
if buildPatches:
# build patch model
patchModel = build_patches(data_patches, 0.00000001, False)
# store the model
model = {}
if buildPatches:
model['patchModel'] = patchModel
model['patchModel']['canvasSize'] = [cropsize[0],cropsize[1]]
if buildScoring:
model['scoring'] = scoring
model['shapeModel'] = {}
model['shapeModel']['eigenVectors'] = eigenVectors.T.tolist()
model['shapeModel']['eigenValues'] = eigenValues
model['shapeModel']['meanShape'] = meanshape.tolist()
model['shapeModel']['numEvalues'] = len(eigenValues)
model['shapeModel']['numPtsPerSample'] = meanshape.shape[0]
model['shapeModel']['nonRegularizedVectors'] = [0]
